{"id":"d2f3629b-52b8-439c-9f0d-b1d8c5591e63","arxiv_id":"1908.08325","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Direction-dependent sensing limits caused by physical barriers generically break the symmetry needed for diffusive macroscopic limits, forcing a hyperbolic description near obstacles.","lead":"A mathematical model of migrating cells now lets the sensing radius shrink when cells face physical barriers, such as dense tissue or crowds. The model shows that near such barriers the cell population obeys a hyperbolic, wave-like equation rather than the usual diffusion equation.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The hyperbolic-limit conclusion rests on an asymptotic expansion that assumes smooth macroscopic variation; sharp barriers make the sensing radius vary on the cell scale, so the global limit (42) is not justified.","rationale":"The paper's central claim, as captured by the reader's strongest_claim, is that near physical limits of migration the leading-order macroscopic velocity U0 does not vanish, making the hyperbolic scaling necessary. The reader's weakest_assumption focuses on the hard-threshold form of the sensing radius in Eqs. (13)-(14). That is a modeling concern, but the more load-bearing issue is internal to the asymptotic argument: the expansion used to derive the macroscopic limit assumes smooth dependence of the turning kernel on the macroscopic variable, while the sharp barriers central to the paper create a sensing radius that is discontinuous and varies on the cell scale. The paper does not provide a boundary-layer analysis or any other justification that the global limit is the hyperbolic equation (42) rather than a parabolic equation with an interface condition. This gap affects the theoretical novelty of the paper, not just its biological assumptions. I therefore recommend keeping the CONDITIONAL verdict, as the paper would need a substantial revision of its scaling analysis or a restatement of the claim to be accepted. My concern is related to but distinct from the reader's; they identified the hard cutoff as the source of the asymmetry, while I question whether the resulting asymmetry survives in the macroscopic limit as claimed.","tokens_in":26675,"tokens_out":11267,"duration_ms":129035,"concrete_test":"Solve the 1D kinetic model of Section 5.1 with a sharp density step ρ=0 for x<0 and ρ>ρth for x>0, with the sensing radius Rmax = εL, for ε = 0.1, 0.05, 0.025. Measure the steady macroscopic velocity U(x) at a fixed macroscopic point x0 ≠ 0, e.g. x0 = -L/2. If U(x0) → 0 as ε → 0, the hyperbolic term is confined to a boundary layer and the global limit is not Eq. (42); if U(x0) tends to a nonzero constant, the hyperbolic claim survives. An analytical matched-asymptotic expansion in the outer variable ξ = x/L and inner variable z = x/Rmax would settle the same question directly.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Section 4 the authors expand the turning kernel as T = T0 + ε T1 + O(ε²) with T0 and T1 depending on the macroscopic variable ξ, and they then conclude from Eqs. (38) and (45)-(46) that near a physical barrier U0 ≠ 0, so the hyperbolic limit (42) is the appropriate one. This expansion assumes T and the sensing radius RM_{S'} are smooth functions of ξ. However, for the sharp barriers that motivate the paper (step-function M in Fig. 2, thresholded densities in Section 5.1), RM_{S'}(t,x,hat v) defined in (13)-(14) is discontinuous in x at the barrier; after rescaling ξ = εx it varies on the fast scale x/ε. Consequently U0 computed from (38) is nonzero only in a boundary layer of width comparable to the sensing radius. In the ε→0 limit this layer collapses, so the bulk macroscopic equation away from the barrier is the parabolic equation (47), with the barrier entering through an interface or no-flux condition rather than through a bulk hyperbolic advection term. The argument in Section 5.1 that η = R̄ρ/lρ > 1 near barriers forbids a diffusive time scale only shows that a single parabolic scaling is not uniform; it does not justify the global hyperbolic equation (42). Without a matched-asymptotic or interface-layer analysis, the central claim that the appropriate macroscopic limit is hyperbolic is not established.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper extends the non-local kinetic model of Loy and Preziosi (2019) to account for physical limits of migration. The key novelty is that the sensing radius in the turning operator is no longer constant but depends on position, sensing direction, and time, through a hard-threshold rule (13)-(14): a cell senses up to the first point where a mechanical cue M exceeds a threshold Mth, plus a small poking depth Δ. The authors derive macroscopic limits of the kinetic transport equation and argue that, near physical barriers, the direction dependence of the sensing radius makes the leading-order macroscopic velocity U0 nonzero, so the appropriate scaling is hyperbolic (Eq. 42), with parabolic scaling possible only away from barriers. The modelling framework is then illustrated with one- and two-dimensional simulations for volume filling, cell-ECM interactions, cell-cell adhesion, and chemotaxis, showing pattern formation, barrier crossing, and aggregation phenomena.","tokens_in":27028,"tokens_out":7330,"duration_ms":72068,"significance":"If the central claim is valid, the paper offers a useful modelling mechanism: physical limits of migration introduce a direction-dependent sensing radius, which breaks the symmetry needed for a purely diffusive macroscopic description. The model is flexible and biologically well motivated, separating polarization from speed sensing and incorporating a hard threshold for sensing. The numerical experiments demonstrate interesting phenomena (e.g., spontaneous pattern formation with wavelength set by the sensing radius, trapping near dense ECM, and the importance of protrusion length for crossing poor-adhesion regions). The paper is clearly written and builds transparently on prior work. However, the theoretical derivation of the macroscopic limit has a gap concerning sharp barriers, which is load-bearing for the paper's central claim that the global appropriate limit is hyperbolic.","major_comments":[{"comment":"The expansion T = T0 + ε T1 + O(ε²) and the subsequent derivation of the hyperbolic limit (42) assume that the turning kernel T and the sensing radius R_{S'}(ξ, v̂) are smooth functions of the macroscopic variable ξ. For the sharp barriers that motivate the model (e.g., the step-like density in Fig. 7 or the thresholded regions in Section 5.1), R_{S'} defined in Eq. (13)-(14) is discontinuous in x at the barrier. Under the rescaling ξ = εx, this discontinuity varies on the fast scale x/ε, so the expansion is not uniform and the limit (42) is not justified as a global macroscopic equation. Away from the barrier the sensing radius is effectively constant and the parabolic limiting equation (47) applies, with the barrier entering through an interface or no-flux condition rather than through a bulk hyperbolic advection term. A matched-asymptotic or boundary-layer analysis is needed to establish the claimed hyperbolic limit, or the claim should be reformulated as a statement about the boundary layer only.","section":"Section 4, Eqs. (28)-(42)"},{"comment":"The argument that η = R̄ρ/lρ ≫ 1 near barriers rules out a diffusive time scale only shows that a single parabolic scaling is not uniform in Ω. It does not imply that the global hyperbolic equation (42) is the correct macroscopic model, because the derivation of U0 in Eq. (60) still requires the Taylor expansion (58) to be legitimate. Near a sharp barrier, lρ is small and the sensing radius Rρ is large compared with lρ, so the nonlocal contribution cannot be approximated by a local gradient term; the macroscopic flux is not a smooth function of ξ. Thus the reasoning does not bridge the gap from 'no uniform diffusive scaling' to 'global hyperbolic equation'.","section":"Section 5.1, Eqs. (62)-(63)"},{"comment":"The statement that condition (41) (U0 = 0) cannot be satisfied when R_{S'} depends on v̂ is too strong. For a cell positioned symmetrically between two identical barriers, Γ_{S'}(ξ, v̂) is even in v̂, and with B constant the integral in Eq. (38) vanishes identically even though R_{S'} is direction-dependent. The paper acknowledges this possibility with 'unless for very peculiar cases', but the exception is not characterized. Since the dichotomy between hyperbolic and parabolic limits is central to the paper, this oversight should be addressed either by giving a precise condition for U0 = 0 or by softening the claim to 'in generic configurations'.","section":"Section 4, Eqs. (38)-(46)"},{"comment":"The numerical simulations are presented without convergence checks, mesh refinement studies, or quantitative validation. The paper makes a theoretical claim about macroscopic limits, so at least one grid-convergence study (for example, for the key simulation in Fig. 7 or Fig. 15) is necessary to rule out numerical artifacts. This is particularly important for the Dirac-delta sensing kernel (Section 5.1), where the observed patterns have a wavelength comparable to the sensing radius and potentially to the grid size. Without such checks, the qualitative conclusions (e.g., that Rρ = Rmaxρ allows densities to exceed the threshold, while the limited-radius model does not) rest on unverified numerical evidence.","section":"Sections 5 and 6"}],"minor_comments":[{"comment":"In Eq. (17) and Eq. (20), the argument of T is written as (x, v, v) instead of (x, v, ˆv); this is a typographical error.","section":"Eq. (17) and (20)"},{"comment":"The notation S′ is used both for the field and for the variable in ψ(v|S′(y)); this dual use is confusing and should be disambiguated, for example by writing the field as S′(x) and the variable as s.","section":"Section 4, after Eq. (43)"},{"comment":"In Eq. (45), the subscript on U0_{S′} is inconsistent with the notation U0_{S,S′} used in Eqs. (38)-(40); this should be harmonized.","section":"Eq. (45)"},{"comment":"The labels in subfigures (e) and (f) are incomplete: 'Rmax_M = 0.2,p' and 'Rmax_M = 0.2,p/ρ' should indicate the time at which the plots are taken and what the color scale represents.","section":"Fig. 12"},{"comment":"The set notation in Eq. (54) is slightly imprecise: [0, Rρ(t,x,ˆv)] = { λ′∈[0,Rmaxρ] | ρ(t, x+λ′ˆv) ≤ ρth } defines an interval of λ′ values, but the right-hand side is a set of admissible λ′; this is acceptable but could be phrased more cleanly.","section":"Section 5.1, Eq. (54)"}],"recommendation":"major_revision","confidential_remarks":"The paper is a preprint from 2019 and appears to have been under consideration for some time. The novelty over Loy and Preziosi (2019) is incremental but real: the direction-dependent sensing radius is a new mechanism with plausible biological grounding. The main obstacle is the macroscopic-limit gap described in the major comments. If the authors can provide a matched-asymptotic treatment of the barrier region or substantially revise the claim to a boundary-layer statement, the paper would be publishable. Otherwise, the central theoretical conclusion is not supported. The simulations are interesting but would benefit from added rigor (convergence checks, parameter tables) before acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper's genuinely useful contribution is the idea that physical barriers limit the sensing radius in a direction-dependent way (Eqs. 13–14), which breaks the parity that a diffusive limit needs. The authors show this in the kinetic setting of their earlier model, and the simulations illustrate the consequences—crowding near thresholds, pattern formation tied to the sensing radius, and the difference between limited and unlimited sensing. That part is solid and worth reading.\n\nThe soft spot is the macroscopic limit. The claim is that because U0 does not vanish near barriers, the appropriate limit is hyperbolic (Eq. 42). But the asymptotic expansion in Section 4 assumes the turning kernel is smooth in the macroscopic variable ξ. For the sharp barriers in their motivating examples (step-function M, thresholded densities), the sensing radius RM is discontinuous at the barrier, and the asymmetry in ΓS′ exists only in a layer of width comparable to the sensing radius. In the ε→0 limit that layer collapses to a surface. So away from the barrier the bulk limit can still be parabolic (Eq. 47), with the barrier entering through an interface condition. The paper's statement that condition (46) fails \"in points close to\" a barrier is true pointwise, but it does not justify a global hyperbolic PDE. The η argument in Section 5.1 shows that a single parabolic scaling is not uniform; it does not establish (42) as the limit.\n\nThis matters because it is the paper's main mathematical takeaway. I don't think it sinks the paper. The model itself is well posed, the mechanics of the turning operator are clear, and the numerical studies are reasonable qualitative illustrations, though they are not convergence-checked or validated against experiments. The absence of a theorem that U0 cannot vanish can be excused in a modeling paper; the global hyperbolic claim should be softened or backed by a boundary-layer analysis.\n\nWho gets value: researchers working on kinetic or nonlocal models of cell migration, especially with volume filling or ECM steric effects. They will find the sensing-radius mechanism useful and the simulations informative. I would send this to a referee rather than desk reject, with the expectation that the limit claim be fixed or reframed.","headline":"A useful modeling extension—direction-dependent sensing radius—but the claimed global hyperbolic macroscopic limit is not proven; the kinetic model and simulations are worth a referee's time.","tokens_in":27481,"tokens_out":6738,"would_cite":true,"duration_ms":64863,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["92C17","35Q92","82C40"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that near physical limits of migration, cells' sensing radius becomes direction-dependent, making the leading-order macroscopic velocity nonzero and forcing a hyperbolic, not diffusive, macroscopic limit.","keywords":["biased cell migration","extracellular matrix","taxis","physical limit of migration","kinetic transport equation","non-local sensing","macroscopic limit","volume filling"],"falsifier":"Compute the zeroth-order macroscopic velocity (38) for the same transition probability but with a smooth sensing kernel that never fully cuts off, for instance $\\gamma_{S'}(\\lambda')=\\exp(-\\lambda'/\\Lambda)$ independent of $M$, while keeping the speed cue unchanged; if $U^0_{S,S'}$ vanishes wherever the speed cue is even in $\\hat v$, then the hard-threshold cutoff is the mechanism forcing the hyperbolic limit. In the laboratory, microfluidic channels with a pore-size gradient could test the predicted direction-dependent speed near a barrier: the model says a cell's measured speed depends on which way it is polarized even with no chemoattractant present.","tokens_in":26487,"feed_emoji":"🧫","tokens_out":9933,"duration_ms":93050,"temperature":0.7,"pith_summary":"The paper argues that at physical limits of migration—overcrowded regions, basal membranes, or extracellular matrix with pores too small for the nucleus—a cell's sensing radius stops being a fixed number and becomes shorter along directions that hit the obstacle. Because the sensing radius now depends on direction, the leading-order macroscopic cell velocity is generically nonzero even when no directional bias is present. The authors therefore conclude that the usual parabolic (diffusive) limit of the kinetic transport equation fails near such barriers, and that the correct macroscopic description is the hyperbolic conservation law of Eq. (42), with diffusive scaling possible only away from the barriers. They support this by deriving the macroscopic limits and by simulating volume filling, cell–ECM interactions, cell–cell adhesion, and chemotaxis. The consequence that matters is practical: continuum models of invasion through heterogeneous tissue should use an advective, not purely diffusive, equation near obstacles.","feed_headline":"Barriers make cell motion drift, not diffuse","feed_subtitle":"A kinetic model shows that direction-dependent sensing near obstacles gives cells a nonzero mean velocity, so hyperbolic limits apply.","key_machinery":"The load-bearing object is the limited sensing radius defined by $R^M(t,x,\\hat v)=\\inf\\{\\lambda\\in[0,R^{\\max}_{S'}):M(t,x+\\lambda\\hat v)>M_{\\mathrm{th}}\\}$ and $R^M_{S'}=\\min\\{R^M+\\Delta,R^{\\max}_{S'}\\}$: a cell integrates information only up to the first point along its polarization direction where the environment is impassable, plus a fixed poking depth (set to zero in the simulations). This replaces the constant sensing radius of the earlier kinetic model and makes the sensing support, and therefore the normalization $\\Gamma_{S'}$, depend on direction. It is this direction-dependent $\\Gamma_{S'}$ that enters the zeroth-order velocity formula and spoils the evenness condition near barriers, forcing the hyperbolic scaling. The polarization factor $B[S]$ and the speed density $\\Psi[S']$ enter the transition probability as independent averages over the direction-dependent interval.","core_discovery":"The discovery is that the direction dependence of the sensing radius, and not the direction dependence of the sensed cue itself, is what breaks the standard diffusive regime. In the model, the sensing distance in direction $\\hat v$ stops at the first point where the mechanical cue $M$ exceeds a threshold $M_{\\mathrm{th}}$, possibly plus a small poking depth $\\Delta$. This makes the effective sensing weight $\\Gamma_{S'}(x,\\hat v)=\\int_0^{R^M_{S'}(x,\\hat v)}\\gamma_{S'}(\\lambda')\\,d\\lambda'$ direction dependent. The zeroth-order macroscopic velocity $U^0_{S,S'}(x)=c(x)\\int_{\\mathbb S^{d-1}}\\Gamma_{S'}(x,\\hat v)B[S]_0(\\hat v)\\bar U^0_{S'}(x|\\hat v)\\,\\hat v\\,d\\hat v$ then lacks the symmetry that would make it vanish, so the condition $U^0_{S,S'}=0$ required for a parabolic limit fails. The paper states this directly: as soon as a physical barrier appears, points close to it develop an asymmetry in the evaluation of the sensing radius that invalidates the evenness condition, and the macroscopic limit is the hyperbolic equation $\\partial_\\tau\\rho+\\nabla\\cdot(\\rho U^0_{S,S'})=0$.","pith_inferences":["An implication the authors leave implicit is that the same parity-breaking mechanism should appear in any model where perception is truncated by obstacles—bacterial chemotaxis in porous media, animal movement in fragmented landscapes, pedestrian flows—making hyperbolic macroscopic limits more common than diffusive ones in confined environments.","The hyperbolic limit implies that density fronts steepen into shocks at barriers, so a stability analysis of the hyperbolic equation could predict the observed pattern wavelength $R^{\\max}$ directly, without resolving the kinetic equation.","Replacing the hard threshold by a smooth, direction-independent attenuation of the signal with distance should weaken the asymmetry; computing $U^0$ under such a kernel would isolate whether the threshold cutoff is the essential ingredient of the conclusion."],"forward_implications":["Near any physical barrier, the continuum description of a cell population should be a hyperbolic conservation law for the density with flux $\\rho U^0_{S,S'}$; diffusive approximations should be used only away from the barrier.","If the sensing radius is not limited at the barrier, simulated cells inside a dense ECM region can sense beyond it and escape; with the threshold-limited radius they stay trapped, matching the biological picture of a nucleus-imposed pore limit.","Cell speed near a barrier depends on polarization direction: a cell at the edge of a crowded region can move outward but not inward, so the macroscopic velocity can be nonzero and directed away from the crowd.","The sensing kernel changes the predicted pattern: a Dirac-delta kernel produces patterns with wavelength of order $R^{\\max}$ and a zig-zag dynamics, while a Heaviside kernel averages the signal and keeps densities smoother and below threshold.","With a nonzero poking depth $\\Delta$, a cell approaching a barrier decelerates gradually rather than stopping abruptly, and a cell within $\\Delta$ of an exit accelerates quickly; a vanishing $\\Delta$ makes the stop at the barrier sharp."],"supporting_citations":[{"why":"Supplies the base kinetic model with non-local sensing of independent cues for polarization and speed, which this paper extends to direction-dependent sensing radii.","marker":"Loy and Preziosi (2019)"},{"why":"Provides the moment conditions and mass conservation used to judge when a diffusive limit of the velocity-jump equation is valid.","marker":"Othmer and Hillen (2000)"},{"why":"Introduces the finite sensing radius and non-local gradient approach for chemotaxis that the non-local sensing terms build on.","marker":"Othmer and Hillen (2002)"},{"why":"Establishes the velocity-jump transport equation framework that the kinetic model uses.","marker":"Othmer et al. (1988)"},{"why":"Supplies the experimental evidence on nucleus-imposed pore-size limits that motivates the threshold $M_{\\mathrm{th}}$ and the physical barrier picture.","marker":"Wolf et al. (2013)"},{"why":"Introduces the velocity-jump process underlying the turning operator.","marker":"Stroock (1974)"}],"fun_headline_variants":["Cell sensing barriers turn diffusion into drift","Asymmetric sensing near walls gives cells a velocity","Kinetic model: barriers make cell drift, not diffuse","Sensing radius asymmetry breaks diffusive scaling","Obstacles change cell motion from diffusion to drift"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the hard threshold in sensing: a cell ignores everything beyond the first point where the mechanical cue exceeds $M_{\\mathrm{th}}$, plus a fixed poking depth that is set to zero in the simulations; if sensing attenuated gradually or could pass through dense regions, the direction asymmetry that produces the nonzero leading-order velocity would no longer be forced.","fun_headline_variants_meta":{"raw":{"variants":["Cell sensing barriers turn diffusion into drift","Asymmetric sensing near walls gives cells a velocity","Kinetic model: barriers make cell drift, not diffuse","Sensing radius asymmetry breaks diffusive scaling","Obstacles change cell motion from diffusion to drift"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00017,"raw_usage":{"total_tokens":1270,"prompt_tokens":946,"completion_tokens":324,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":562,"completion_tokens_details":{"reasoning_tokens":253}},"tokens_in":562,"tokens_out":324,"duration_ms":4138,"temperature":1.0,"reasoning_tokens":253,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:41:23.592576+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the zeroth-order macroscopic velocity (38) for the same transition probability but with a smooth sensing kernel that never fully cuts off, for instance $\\gamma_{S'}(\\lambda')=\\exp(-\\lambda'/\\Lambda)$ independent of $M$, while keeping the speed cue unchanged; if $U^0_{S,S'}$ vanishes wherever the speed cue is even in $\\hat v$, then the hard-threshold cutoff is the mechanism forcing the hyperbolic limit. In the laboratory, microfluidic channels with a pore-size gradient could test the predicted direction-dependent speed near a barrier: the model says a cell's measured speed depends on which way it is polarized even with no chemoattractant present.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the moment conditions and mass conservation used to judge when a diffusive limit of the velocity-jump equation is valid."},{"cited_title":"and Hillen, T","cited_arxiv_id":null,"evidence_quote":"Introduces the finite sensing radius and non-local gradient approach for chemotaxis that the non-local sensing terms build on."},{"cited_title":"G., Dunbar, S","cited_arxiv_id":null,"evidence_quote":"Establishes the velocity-jump transport equation framework that the kinetic model uses."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the experimental evidence on nucleus-imposed pore-size limits that motivates the threshold $M_{\\mathrm{th}}$ and the physical barrier picture."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the velocity-jump process underlying the turning operator."}],"review_version":1}